Making ScaleConf more technical

gistfile1.txt

Abz, been thinking about this.

Would workshops fill your requirements?

If we had a day before the main conference where we had two streams of
four workshops, about 90 minutes each...  Some expert from Percona
talking about mysql optimization, somebody teaching the basics of
puppet or chef, somebody demonstrating vagrant or capistrano...

Would that provide the technical benefit of the conference to you?
Talks that are technical get boring quite quickly because slides don't
convey to people what is actually going on.  I'm not sure how to get
around this.

Anyway, your thoughts would be welcome.

Cheers,
-Jonathan

===========================

Hi Jonathan                                      >@2012.01.30_23:36:20_+0200

Yes, workshops might do the trick.

However, I think it is crucial to find out what people actually want to
workshop.

For instance, topics of interest to me:

 a) I would love to find out more about large scale distributed file stores
   e.g. a technical talk/workshop about Amazon's EBS implementation (and
   their new storage plans), GlusterFS (or equivalents), Lustre (with HA?),
   HA NFS, large scale deployments of cluster file systems, etc.

 b) Distributed file stores for large binary objects, e.g. S3 or SWIFT
   implementations (and different strategies: eventual consitency vs
   hashed, etc)

 c) How companies have solved their scaling problems for mission critical
   applications (e.g. billing for Vodafone) where the data needs to be
   accurate at all times, i.e. where transactions are of utmost importance,
   most caching strategies fail, etc

 d) Mysql optimization and experience in large scale deployments,
   geographically distributed Mysql clusters (or replication), etc. Esp
   where people have already scaled up (SSD, optimized Mysql, high speed
   interconnects, lots of memory) and out (clusters and/or replication) as
   well as experience with different read/write loads and different tables,
   e.g. we have a few cases where InnoDB sucks compared to MyISAM due to
   high write loads, interesting ways in which we've had to solve
   transaction serialization and deadlocks with InnoDB for complex high
   concurrency transactions, etc

 e) Experience with different cloud management platforms, e.g. Eucalyptus,
   Openstack, Nimbula, CloudStack, Convirture, etc and issues experienced
   with them.

 f) How people have solved their bandwidth billing problems on shared
   infrastructure, e.g. DaaS where you might have multiple customers
   sharing single IP address, hosted Exchange/Zimbra, etc.

 g) Large scale desktop, thin client, and mobile (tables, phones)
   deployments and management thereof.

 h) Scaling problems with different languages and frameworks e.g.
   Ruby/Rails, Python/Django, PHP/Zend/Cake/CI/Drupal, etc. For instance it
   would be fascinating to hear somebody from Facebook talk about how they
   scaled PHP in their environment.

 i) How people manage their transactions in multi-db environments such as
   NoSQL+SQL, splitting their data into DB per customer, etc. Our
   experience doing this was akin to masturbating with a cheese grater.
   Rather amuzing, but mostly painful...

and so on (:

I suspect that most people have already figured out how to work with
config/image management tools such as Chef/Puppet/Capistrano, half the world
has now played with NoSQL implementations such as Couch/Redis/Mongo, etc

The real benefit of such a workshop should be to listen to or debate with
experts that have solved or are solving the difficult problems which most of
us has not solved yet or are hacking around (e.g. using SANs instead of
distributed file storage because it is too difficult to do).